CroNER is a tool for recognizing and classifying named entities in natural language texts in Croatian. CroNER recognizes nine different classes of named entities (see Specification & Features below). Although developed for Croatian, CroNER can successfully be applied to texts in closely related languages such as Slovenian, Serbian, Bosnian, Montenegrian, and Macedonian. CroNER uses a powerful sequence labelling model, namely Conditional Random Fields (CRF), coupled with smartly engineered Named Entity Recognition features. Some of the features specifically account for morphological complexity typical for Slavic languages.

• State-of-the-art named entity tagging for Croatian (accuracy above 90%, a top-performance among Slavic languages)
• Recognizes named entities of nine different types: Persons (and separately Personal Possessives), Locations, Organizations, Ethnics, Date and Time, Money, and Percentage expressions
• Accurate named entity tagging for closely related languages: Serbian (latin), Bosnian, Montenegrian, Slovenian, and Macedonian
• Available as a standalone desktop tool, web tool, or via a web service
• Fast named entity tagging (processing at 2000 words per second)
• Inline and stand-off output annotations

CroNER is primarily a commercial tool. However, access to CroNER Web application is freely available for research purposes. If you are interested in using CroNER either for research purposes or commercially, please drop us an email at info@takelab.hr.